Company Sentiment Done Differently
For many years, the ability to decipher the sentiment of a piece of text, a document, or an entire corpus of data has proven valuable. It enables organizations to understand the tone and tenor of content about themselves, their competitors, or the market, and then to make strategic decisions as a result.
But a harder problem to solve has involved entity-level sentiment. Said another way: Understanding more than just the overall sentiment of a document or news article or social media post; but understanding the sentiment about a particular person, place, or thing mentioned in the content. Because, surely, the classification of a single document as positive or negative has a lot to do with answering the question: positive or negative for whom?
Answering this question for companies in particular has clear and important value. Today, scores of institutions in the financial sector work hard to understand company sentiment because it tells them a lot about companies that they’re interested in.
Among many use cases, these institutions use company sentiment to understand the tone of news coverage about companies. They use sentiment signals for alpha generation, which involves developing investment and portfolio management strategies with returns that exceed the market as a whole. Sentiment is used in the financial sector to monitor a portfolio’s performance and risk in order to ensure that investor interests are protected. It’s also used to stay abreast of or to anticipate changes in regulatory policy or interpretation.
We know that understanding company sentiment is a powerful capability because many of our clients use Finch AI’s sentiment approach for the same purposes – and they tell us our approach is meaningfully different.
In addition to identifying a piece of content as positive, negative, or neutral, Finch AI returns a numeric score that indicates how positive, negative, or neutral the item is. We can do this because we go beyond just keywords to understand sentiment in text. Our models were trained on a large and diverse corpus of news, conversational and narrative datasets to capture nuances in language that other products cannot. We then go a level deeper and deploy these same sentiment models to decipher sentiment at the entity level – rather than just at the sentence or document level. We apply this context-based approach on large documents, rich with entities.
Over time, Finch AI has built a massive knowledge base of millions of companies from around the world. We use this knowledgebase in combination with our topic-based approach to assigning sentiment and our other entity intelligence capabilities to give users access to information about every entity mentioned in a given document. Users get an overall sentiment score for the entity in question, a sentiment score per-mention, and rich metadata about companies of interest.
We also return topics and key phrases that offer additional context about an entity or a document. For each company mention, we return a specific business topic. We have identified two dozen business and financial sentiment topics that help examine and infer context that contributes to an understanding of sentiment about a company. These topics help us correctly interpret things such as sales that “explode” which is a positive thing, whereas in nearly every other context, something exploding would be decidedly not positive. These topics and key phrases also assist in disambiguating an entity’s identity – meaning correctly deciphering between identically named entities of the same type by using the context surrounding that entity in a given document.
Our approach to company sentiment is helping customers develop deeper insights and understanding of the environments in which they operate – and those insights are informing strategy and putting distance between our customers and their competitors.
We look forward to continuing to leverage AI and its emerging use cases and capabilities to further extend our own language and entity intelligence offerings, including enhancing our already novel approach to sentiment. We’re also excited about being a part of the continued evolution of AI and enabling customers to tackle additional real-time, enterprise-scale data challenges.
To learn more about what we’re doing and how it’s impacting organizations of all types, please visit www.finchai.com.